# Maximum likelihood cross-validation KDE question

To find the optimal bandwidth $$h$$ in maximum likelihood cross-validation for kernel density estimation (KDE), is there a better way to find the optimal $$h$$ than just gradient ascent?